Tracking Issue: out-of-tree modules failing to build on Linux 6.13

[SigmaSquadron]

Problem Description

Since the Linux 6.13 upgrade, several out of tree modules have been failing to build with similar logs. Multiple issues have been filed thus far, and this is a helper issue to centralise the individual module reports.

Since the logs seem to be the same, it stands to reason that the cause of the build failure is the same, so hopefully there's a single solution for all of these issues. The PR author that will fix this issue can then easily link back to all of the individual issues.

Issue List

These issues fail with mkdir: cannot create directory '.tmp_##': Permission denied

• openrazer: fails to build, permission denied #375282
• Build failure: v4l2loopback #375439
• Build failure: linuxPackages_6_13.xpadneo #375591
• Build failure: linuxKernel.packages.linux_6_13.acpi_call.aarch64-linux #375641
• Build failure: gcadapter-oc-kmod #375731
• Build failure: vhba #375737
• Build failure: linuxKernel.packages.linux_6_13.evdi #375917

Issue List (Other)

These issues fail with different logs, but all seem to be related to an incorrect working directory.

• Build failure: nvidia-x11-565.77-6.13 #375730
• Build failure: linuxKernel.packages.linux_6_13.broadcom_sta #375801
• Build failure: linuxKernel.packages.linux_6_13.cpupower #375865

---

Add a 👍 reaction to issues you find important.

[Bot-wxt1221]

@SigmaSquadron I'm strange to see that set buildRoot will fix the problem. I'm not sure why it was used. But now it is empty. I fix it by export buildRoot=.

[misuzu]

cc @K900

[nixos-discourse]

This issue has been mentioned on NixOS Discourse. There might be relevant details there:

https://discourse.nixos.org/t/nvidia-565-77-wont-work-in-kernel-6-13/59234/5

[raj-magesh]

Not sure if I should open a separate issue, but in case anyone's wondering why, ddcci-driver-6.13-0.4.5-unstable-2024-09-26.drv also fails to build with similar errors.

[fpletz]

Adding "KBUILD_OUTPUT=${kernel.dev}/lib/modules/${kernel.modDirVersion}/build" to makeFlags in the affected kernel modules also fixes the issue. Adding this to every module might be the way to go but we might need to guard it for 6.13+ only. Maybe we should add this to the kernel's makeFlags since they are generally added via kernel.makeFlags to the makeFlags of out of tree modules.

[K900]

FWIW I'm not sure that's the right solution, though the whole module build system is already a big mess upstream.

[K900]

Thinking about it more, we should probably add a separate moduleMakeFlags or similar variable that sets things specifically for out of tree modules, and use that.